added InputList

src/TensorFlowNET.Core/Eager/Execute.cs

@@ -6,12 +6,9 @@ namespace Tensorflow.Eager
 {
     public class Execute
     {
-        public void record_gradient(string op_name, Tensor[] inputs, Dictionary<string, object> attrs, Tensor[] results, string name = "")
+        public void record_gradient(string op_name, InputList inputs, Dictionary<string, object> attrs, Tensor[] results, string name = "")
         {
-            if (inputs == null)
-                inputs = new Tensor[0];
-
-            pywrap_tfe_src.RecordGradient(op_name, inputs, attrs, results, name);
+            pywrap_tfe_src.RecordGradient(op_name, inputs._inputs, attrs, results, name);
         }
     }
 }

src/TensorFlowNET.Core/Eager/pywrap_tfe_src.cs

@@ -25,6 +25,9 @@ namespace Tensorflow.Eager
                 }
             }
             if (!should_record) return;
+
+            var op_outputs = results;
+            var op_inputs = inputs;
         }
     }
 }

src/TensorFlowNET.Core/Operations/InputList.cs

@@ -0,0 +1,16 @@
+using System;
+using System.Collections.Generic;
+using System.Text;
+
+namespace Tensorflow
+{
+    public class InputList
+    {
+        public Tensor[] _inputs;
+
+        public InputList(Tensor[] inputs)
+        {
+            _inputs = inputs;
+        }
+    }
+}

src/TensorFlowNET.Core/Operations/Operation.cs

@@ -101,21 +101,23 @@ namespace Tensorflow
             }
         }
 
-        private Tensor[] _inputs;
-        public Tensor[] inputs
+        private InputList _inputs;
+        public InputList inputs
         {
             get
             {
                 if(_inputs == null)
                 {
-                    _inputs = new Tensor[NumInputs];
+                    var retval = new Tensor[NumInputs];
 
                     for (int i = 0; i < NumInputs; i++)
                     {
                         var tf_outpus = Input(i);
                         var op = new Operation(tf_outpus.oper);
-                        _inputs[i] = op.outputs[tf_outpus.index];
+                        retval[i] = op.outputs[tf_outpus.index];
                     }
+
+                    _inputs = new InputList(retval);
                 }
 
                 return _inputs;